Detecting Embankment Risks with Satellites and AI: New Feature Added to SateAIs

The Fusion of Satellite Data and AI Opens New Horizons for Risk Management

On April 22, 2026, Space Shift Inc. announced the launch of a new feature, the “Embankment Detection AI,” within its satellite data analysis brand, SateAIs. This groundbreaking service uses AI to analyze time-series image data collected from multiple satellites, automatically identifying areas with potential embankments based on surface changes.

Embankments refer to artificially created land areas built using excavated soil or waste materials for purposes such as constructing roads, residential areas, or parks. During Japan’s rapid urbanization in the post-war economic boom, embankments were extensively used, and today there are estimated to be over 50,000 such sites nationwide. However, many embankments lack accurate records, leaving their locations and scales largely unknown. The 2021 landslide disaster in Atami City, Shizuoka Prefecture, was later found to have been caused by a large embankment, highlighting the critical importance of proper embankment management.

Traditionally, identifying embankments relied on on-site surveys and historical document reviews, which were time-consuming, costly, and prone to errors due to incomplete records. The “Embankment Detection AI” by SateAIs aims to tackle these challenges through technological innovation.

How It Works: The AI’s Eye for “Change”

The core of this AI lies in its ability to detect subtle “changes” on the surface, not from a single satellite image, but by comparing multiple images taken of the same location over several years. Areas where embankments have been constructed typically show distinct patterns that differ from natural landscapes, such as abrupt changes in elevation immediately after construction, followed by gradual subsidence and vegetation changes over time.

The deep learning model developed by SateAIs has been trained to recognize these specific patterns in time-series data. For instance, it can identify areas where the ground level has rapidly increased within a short period and then gradually subsided, or where uniformly sloped terrains have formed, deviating from the surrounding natural topography. By incorporating radar satellite data, which captures three-dimensional surface shapes, the system can perform analysis even in cloudy conditions or at night.

“The AI provides results as probability scores. For instance, it might indicate an 80% likelihood of an area being an embankment. These results can then be cross-referenced with historical satellite images and relevant administrative data to prioritize field investigations efficiently,” explained a Space Shift technical representative. In this way, the AI helps optimize the allocation of human resources by narrowing down areas of interest.

Industry Impact: From Disaster Prevention to Urban Renewal

The applications of this technology are vast. First and foremost, it is expected to significantly contribute to disaster prevention administration by municipalities and Japan’s Ministry of Land, Infrastructure, Transport, and Tourism. Embankments are at high risk of causing landslides or mudflows when subjected to heavy rainfall or other forces, and identifying them beforehand is critical for drafting evacuation plans and reinforcing infrastructure. By visualizing previously “invisible risks,” this technology can greatly enhance disaster preparedness.

Secondly, it has potential applications in real estate development and urban planning. For large-scale housing or commercial projects, this AI can quickly assess land safety, enabling developers to mitigate risks during the planning stage and avoid unnecessary costs. Moreover, adding embankment risk layers to 3D maps of entire cities could aid in crafting smarter urban regeneration strategies.

Thirdly, the technology could be utilized in insurance and finance. Insurance companies could use it to precisely evaluate risks associated with specific properties or areas, allowing for accurate premium calculations and underwriting decisions. For investors, access to objective data on real estate risks could improve decision-making processes.

Challenges and Future Outlook: Improving Accuracy and Integrating Data

Of course, challenges remain. The detection accuracy of the AI depends heavily on the quality and quantity of the satellite data it is trained on. To achieve comprehensive and highly accurate detection across Japan’s diverse landscapes and soil types, further training with a wider range of data and on-site verification (collection of ground truth data) will be essential. Additionally, integrating the AI with older, non-digitized administrative records will be critical for unlocking its full potential.

The embankment detection technology also holds promise for detecting other types of surface changes. For example, it could be used to identify illegal waste dumping sites, monitor environmental changes caused by mining activities, track deforestation, and even contribute to broader environmental monitoring efforts. Space Shift has expressed its ambition to evolve SateAIs into a comprehensive platform for multi-faceted surface change analysis in the future.

The integration of satellite data and AI is beginning to provide us with innovative tools to better understand our planet and coexist with risks. SateAIs’ “Embankment Detection AI” serves as a concrete example of how technology can be harnessed to address pressing challenges, offering valuable insights for future advancements in this field.

FAQ

Q: How does the Embankment Detection AI identify areas with potential embankments?
A: The AI compares multiple satellite images of the same location taken over several years. It learns to detect unique surface change patterns associated with embankments, such as sudden elevation changes or subsequent subsidence, using deep learning techniques.

Q: How can this technology assist in disaster prevention efforts?
A: By visualizing previously hard-to-detect embankment risks, municipalities can identify hazardous areas in advance and prioritize measures such as preparing evacuation routes or reinforcing ground stability. This contributes to mitigating damage during disasters like heavy rain.

Q: How accurate is the satellite data analysis, and what are the challenges?
A: The accuracy depends on the quality and variety of training data. While current results show high precision, achieving nationwide coverage at a high level of accuracy requires further training with diverse data and on-site verification. Collaborating with older administrative records is also crucial for enhancing precision.